A Fortran Code for Null Geodesic Solutions in

the Lemaı̂tre-Tolman-Bondi Spacetime

Marcelo B. Ribeiro 1,2

arXiv:gr-qc/0205095v1 22 May 2002

Vatican Observatory Group, Steward Observatory, University of Arizona, Tucson,

AZ 85721, USA; e-mail: mribeiro@as.arizona.edu

Abstract

This paper describes the Fortran 77 code SIMU, version 1.1, designed for numerical
simulations of observational relations along the past null geodesic in the Lemaı̂tre-
Tolman-Bondi (LTB) spacetime. SIMU aims at finding scale invariant solutions of
the average density, but due to its full modularity it can be easily adapted to any
application which requires LTB’s null geodesic solutions. In version 1.1 the numerical
output can be read by the GNUPLOT plotting package to produce a fully graphical
output, although other plotting routines can be easily adapted. Details of the code’s
subroutines are discussed, and an example of its output is shown.

Key words: cosmology, numerical simulation; solution of equations, numerical

relativity
PACS: 98.80.-k, 02.60.Cb, 04.25.Dm

1 Introduction

It is a basic result of relativity theory that light rays follow null geodesics in
curved 4-dimensional spacetimes, and this means that if one is pursuing cos-
mological applications of solutions of Einstein’s field equations with the aim
of comparing the model’s theoretical predictions with the actual data pro-
duced by astronomical observations, one requires null geodesic solutions for
the chosen metric. In fact, for cosmological applications one only needs the
past null geodesics, as what we observe today are events which occurred in
the past. Nevertheless, a very basic problem arising in cosmology, and which
hinders comparison between theory and observations in other models than
1 On leave from Physics Institute, University of Brazil – UFRJ, Rio de Janeiro.
2 Supported by Brazil’s CAPES Foundation.
the standard Friedmann-Lemaı̂tre-Robertson-Walker (FLRW), stems from the
fact that as soon as we depart from the simple FLRW models largely used by
observational cosmologists, the task of finding null geodesic solutions quickly
becomes an intractable analytical problem. In fact, this situation is not fully
appreciated by many of those dealing with cosmological modelling of astro-
nomical data, inasmuch as the simple FLRW models largely applied in ob-
servational cosmology is the exception, since it has a very simple analytical
solution for the null geodesic equation.

The Lemaı̂tre-Tolman-Bondi (LTB) spacetime is the most general spherically

symmetric dust solution of Einstein’s field equations, having the FLRW mod-
els as special sub-cases [1–3]. By being a spatially inhomogeneous model, that
is, where ρ = ρ(r, t), it allows us to study scenarios which do not have FLRW’s
spatial homogeneity assumed a priori. Due to this, it is the most widely used
cosmological model after the standard FLRW, having been applied in a wide
range of cosmological problems, from microwave background radiation studies
to hierarchical (fractal) modelling, just to quote a few (see [3], and references
therein, for a large number of applications of LTB models in cosmology).
However, as is the case with almost all non-standard cosmologies, while it is
possible to solve analytically its Einstein’s field equations, LTB’s null geodesic
equation remains an intractable analytical problem, unless we rewrite its ge-
ometry in terms of the so-called “observational coordinates”. That, however,
has the handicap of adding a great deal of mathematical complexity to the
problem, since such an approach requires complex mathematical calculations
in curved spacetimes [4], something which may not be required in all LTB
applications. Therefore, due to such a wide range of application, a code which
produces numerical solutions of LTB null geodesics is desirable.

In this paper I will describe such a code. It deals with LTB geometry in its
full generality, being applicable to each of its special sub-cases, parabolic,
hyperbolic and elliptic [5], either separately or together in a single problem
encompassing all sub-cases, if desired. It was originally designed to find scale
invariant solutions of the average density by means of numerical simulations
[6,7], but as we shall see below, without changing its core null geodesic calcu-
lations the code can be easily adapted to do much more than this, due to its
modularity. The original results obtained by this code were recently analyti-
cally confirmed, and extended, by other authors [8] by means of the observa-
tional coordinates approach mentioned above. The confirmation of the results
of [7] by [8] adds then further reliability to the numerical results obtained with
the code that will be described below. Besides, those two approaches, numer-
ical and analytical, when applied to a difficult problem such as null geodesic
solutions in LTB spacetime, will in fact complement each other, rather than
exclude one another. In §2 I will describe version 1.1 of the code, and in §3 the
input and output will be discussed. The results are summarized in §4, where
I also indicate where the code can be obtained.

2
2 Description of SIMU 1.1

SIMU solves simultaneously two ordinary differential equations (ODE’s) by

means of the fourth order Runge-Kutta method with adaptive stepsize con-
trol. One differential equation is required for the null geodesic, while the other
is needed for obtaining the redshift in this cosmology [6]. The subroutines that
carry out the numerical integration are from [9], and some minor changes nec-
essary for taking the results back to the main program were made, but without
changing the actual numerical implementation of the Runge-Kutta method.
A full description of the LTB model and notation as used in SIMU is given in
[6]. The numerical simulations take advantage of the fact that LTB spacetime
geometry has three unknown functions, 3 F (r), f (r), β(r), respectively repre-
senting the amount of gravitational mass within the radius coordinate r, the
overall curvature and dynamics of the model, and the time elapsed since the
big bang for each observer located at particular values of r [2,3,5]. The sim-
ulations then take advantage of this freedom, as one starts by choosing these
functions, run the program and analyse the results, concluding then whether
or not the simulation was successful, and if not choose another set of three
functions (see [7] for a very detailed explanation of this procedure).

The code is implemented in double precision, and it has a built-in method-

ology for checking the possible catastrophic accumulation of round-off errors.
That is done by monitoring the energy equation, as derived from Einstein’s
field equations, and its derivative, since, by theory, both must remain un-
changed throughout the integration [7]. In addition, to check for accuracy and
stability, after the integration the program runs in reverse, that is, it takes
the final result and uses it as initial conditions to get another result that can
be compared with the original initial condition. The results described in [7]
showed that double precision is enough for maintaining the desired accuracy
required by the problem under study. SIMU also calculates the errors associ-
ated with each observational quantity evaluated, as well as the propagated
errors. This is done by taking the accuracy given by the adaptive method and
using this value as the input in the standard propagation error equations. All
those details are fully explained in the initial documentation of the code.

As mentioned above, the code uses some subroutines from [9], namely the
fourth order Runge-Kutta integration with adaptive stepsize control, the root
finding algorithm for the transcendental equations appearing in LTB’s ellip-
tical and hyperbolic sub-cases [6], the extended trapezoidal rule quadrature
algorithm, as well as Simpson’s rule to the desired accuracy. However, apart
from those, the remaining subroutines are new, as well as the actual way in
which the whole set of subroutines were linked to each other and organized.

3 Actually two, as a third can be eliminated by a coordinate transformation.

3
In addition, the code is extensively commented and widely documented, also
having a structure chart showing the dependence among the subroutines and
functions. This detailed documentation is added in order to help readers who
might be interested in implementing SIMU in their own computer environ-
ments, or changing it to suit their specific applications.

It must be mentioned that there are more “state of the art” ways for finding
numerical solutions of ODE’s than using the adaptive 4th order Runge-Kutta
method. The ODEPACK subroutine package is an example [10–12]. Never-
theless, although the chosen ODE integrator for SIMU may have already been
superseded by better methods, it does solve the proposed problem, and at the
desired accuracy. Inasmuch as SIMU is a tested code, with its results having
already been analytically confirmed by other authors [8], there is no need to
change to another ODE integrator at SIMU’s current version. The reader must,
however, be aware that such a change might be required for other applications
of SIMU to LTB spacetime.

The code uses only standard Fortran 77 commands to avoid possible com-
pilation problems with machine dependent commands that are not available
universally. If a potential user is only interested in the core of the program, that
is, the part which integrates the null geodesic and calculates the observational
relations, he or she can simply remove its “tail”, where the observational rela-
tions are manipulated according to the aims of the theory studied in [6,7], and
replace it with something else. This “tail” consists of the two last branches
of the structure chart headed by the subroutines FIT and OUTPUT, and are
easily spotted in the MAIN. Finally, the functions f (r), F (r), β(r) and their
derivatives appear in six subroutines at the end of the list so that they can be
compiled separately.

The previous version 1.0 (formerly SIMU 5d; see [13]) had the PGPLOT plotting
routines merged into the second half of the OUTPUT subroutine, while version
1.1 removed them and piped the numerical results into eleven independent
files, v1 to v11, that are then externally read by the GNUPLOT plotting package
in order to produce a graphical output of the results. This is the only differ-
ence between versions 1.0 and 1.1, meaning that in version 1.1 the graphical
output is completely independent from the code itself. Making this change
was justified on two grounds: (i) the GNUPLOT package has greater plotting
capabilities, even allowing LATEX output formats that can be included directly
in the figure environment of any LATEX macro. So, producing plots with the
results is done by feeding the numerical tables into GNUPLOT itself by means
of a script file written for, say, a linux environment; (ii) this feature of hav-
ing the results outputted into numerical files brings flexibility to the code,
as its results can be independently used by another program, if so desired or
demanded by another application of the LTB spacetime.

4
3 Input and Output

The actual implementation for the original problem advanced in [6] was de-
scribed at length in [7], where one can find detailed plots with the results and
analysis of the various simulations, 4 as well as detailed explanations of initial
conditions and functions used in each simulation. So, in here I shall limit the
discussion to the code itself, without dealing with any specific application.

As mentioned above, we require a script file for running a simulation and

immediately producing a graphical output. An example of such a script in
linux is given as follows:

g77 simu1.1.for
a.out < in.simu
gnuplot < in.gnu
latex plots.tex
xdvi plots.dvi

Here the file in.simu contains the maximum value for the radius coordinate r
for the integration, while in.gnu is a short GNUPLOT script file for inputting the
v1 to v11 files and producing LATEX files (see appendix A). Finally plots.tex
is a file for running the .tex LATEX files produced by GNUPLOT in order for the
results be available for analysis directly on the screen.

The details of each simulations, as well as accuracy monitoring are given in an

outputted file called s1. Appendix B shows the results of a simulation using the
value 1.5 for in.simu, and the three functions given as follows: f (r) = cosh(r),
F (r) = sinh3 (r), β(r) = 0.7.

4 Conclusion

In this paper I have described the Fortran 77 code SIMU version 1.1 for calcu-
lating solutions of the null geodesics equations in the Lemaı̂tre-Tolman-Bondi
spacetime geometry. I have discussed the details of the code, input and output,
as well as the ways in which it can be possibly modified for other cosmological
applications of this geometry. The code is available for download at [14].

4 These results were summarized in [2].

5
Acknowledgements

I am grateful to William R. Stoeger for reading the original manuscript, New-

ton José de Moura Jr. for his assistance with the GNUPLOT package, and the
referee for helpful comments. I also thank the Vatican Observatory Research
Group for their kind hospitality while writing this paper.

References

[1] M. B. Ribeiro, Astrophys. J. 395 (1992) 29

[2] M. B. Ribeiro, in: D. Hobill, A. Burd and A. Coley, eds., Deterministic Chaos
in General Relativity (New York, Plenum, 1994) 269-296

[3] A. Krasiński, Inhomogeneous Cosmological Models (Cambridge University

Press, 1997)

[4] R. Marteens, N. P. Humphreys, D. R. Matravers and W. R. Stoeger, Class.

Quantum Grav. 13 (1996) 253; corrigenda 13 (1996) 1689; M. E. Araújo and
W. R. Stoeger, Phys. Rev. D 60 (1999) 104020

[5] W. B. Bonnor, Monthly Not. Royal Astron. Soc. 167 (1974) 55

[6] M. B. Ribeiro, Astrophys. J. 388 (1992) 1

[7] M. B. Ribeiro, Astrophys. J. 415 (1993) 469

[8] N. P. Humphreys, D. R. Matravers and R. Marteens, Class. Quantum Grav. 15

(1998) 3041, gr-qc/9804025

[9] W. H. Press, B. P. Flannery, S. A. Teukolsky and W. T. Vetterling, Numerical

Recipes, (Cambridge University Press, 1986)

[10] A. C. Hindmarsh, ‘ODEPACK, a Systematized Collection of ODE Solvers’, in:

R. S. Stepleman et al., eds., Scientific Computing (North-Holland, 1983) 55-64

[11] L. R. Petzold, SIAM J. Sci. Sta. Comput. 4 (1983) 136

[12] L. R. Petzold and A. C. Hindmarsh, ODEPACK subroutine package available

from netlib at http://www.netlib.org/odepack/

[13] M. B. Ribeiro, On Modelling a Relativistic Hierarchical (Fractal) Cosmology by

Tolman’s Spacetime, PhD thesis, (Queen Mary and Westfield College, University
of London, 1992); http://www.if.ufrj.br/~mbr/publi.html

[14] SIMU 1.1 can be downloaded from http://www.if.ufrj.br/~mbr/codes or

http://arXiv.org/abs/gr-qc/0205

6
A in.gnu Script

set terminal latex

set output ’v1.tex’
set size 1,1
set xlabel "$ \log d l (Gpc)$"
set ylabel "$ \log \rho$"
set nokey
plot ’v1’ with points 1 8, ’v11’ with lines 1
set output ’v2.tex’
set xlabel "$d l (Gpc)$"
set ylabel "$ \rho$"
plot ’v2’ with points 1 8
set output ’v3.tex’
set xlabel "$z$"
set ylabel "$d l$ \\ $(Gpc)$"
plot ’v3’ with points 1 8
set output ’v4.tex’
set xlabel "$z$"
set ylabel "$N c$"
plot ’v4’ with points 1 8
set output ’v5.tex’
set xlabel "$r$"
set ylabel "$Field$ \\ $EE$"
set yrange [-10**-5:10**-5]
plot ’v5’ with points 1 8
set output ’v6.tex’
set xlabel "$r$"
set ylabel "$Deriv.$ \\ $Field$ \\ $EEDSH$"
set yrange [-10**-5:10**-5]
plot ’v6’ with points 1 8
set output ’v7.tex’
set title "$Center \rightarrow Border$"
set xlabel "$r$"
set ylabel "$t$"
set autoscale y
plot ’v7’ with points 1 8
set output ’v8.tex’
set title "$Border \rightarrow Center$"
set xlabel "$r$"
set ylabel "$t$"
plot ’v8’ with points 1 8
set output ’v9.tex’
set title "$Center \rightarrow Border$"

7
set xlabel "$r$"
set ylabel "$I$"
plot ’v9’ with points 1 8
set output ’v10.tex’
set title "$Border \rightarrow Center$"
set xlabel "$r$"
set ylabel "$I$"
plot ’v10’ with points 1 8
quit

8
B Plots of a Single Simulation

-2.5
-3 c c c
c
-3.5 c
-4 c
log ρ c
-4.5 c
-5 c

-5.5 cc

-6
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
log dl (Gpc)

0.0008 c
c
0.0007
0.0006 c
0.0005
ρ 0.0004 c
0.0003
0.0002 c

0.0001 c
c
c
0 c c c c
0 10 20 30 40 50 60
dl (Gpc)

9
 60

50 c
c
40
dl
(Gpc) 30 c
20
c
10 c
c
c
0 ccc
c
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
z

3e+11
c
2.5e+11 c

2e+11

N1.5e+11
c
c
1e+11
c
5e+10
c
c
0 ccc c
c
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
z

10
 1e-05

5e-06

F ield 0 cc c c c c c c c c c
EE

-5e-06

-1e-05
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
r

1e-05

5e-06

Deriv.
F ield 0 cc c c c c c c c c c
EEDSH

-5e-06

-1e-05
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
r

11
 Center → Border
0 cc
c
-0.5
c
-1
c
-1.5
t c
-2
c
-2.5 c
c
-3 c c

-3.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
r

Border → Center
0 cccc
-0.5
c
-1
-1.5 c
t
-2 c

-2.5 c
c
-3 c cc

-3.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
r

12
 Center → Border
0.9
c c
0.8 c
0.7 c
0.6 c
0.5 c
I
0.4 c
0.3
0.2 c

0.1 c
c
0 c
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
r

Border → Center
0.9
0.8 c cc
c
0.7
c
0.6
c
0.5
I
0.4 c
0.3
0.2 c

0.1
ccc
0 c
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
r

13